© January 27, 2006 Dr. Lynn Fuller, Motorola Professor Simulation Page 1 Rochester Institute of...
-
date post
15-Jan-2016 -
Category
Documents
-
view
219 -
download
0
Transcript of © January 27, 2006 Dr. Lynn Fuller, Motorola Professor Simulation Page 1 Rochester Institute of...
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 1
Rochester Institute of TechnologyMicroelectronic Engineering
ROCHESTER INSTITUTE OF TECHNOLOGYMICROELECTRONIC ENGINEERING
Computer Simulation of Factory Performance
Katie McConky ISE Graduate Student
Rochester Institute of Technology 82 Lomb Memorial Drive Rochester, NY 14623-5604
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 2
Rochester Institute of TechnologyMicroelectronic Engineering
OUTLINE
Overview of ModelingVerification and Validation of ModelsModel AccuracyStatistical AnalysisAutoSched AP – Brooks AutomationAutoMod– Brooks AutomationExample Simulation Problems
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 3
Rochester Institute of TechnologyMicroelectronic Engineering
Overview of Modeling
•Semiconductor Fabs are extremely complex
•Reentrant Manufacturing
•High variability
•Long Process Routes
•Complicated batching and sequencing criteria
•Simulation modeling can take everything into account in order to answer questions about the fab.
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 4
Rochester Institute of TechnologyMicroelectronic Engineering
Verification and Validation
Verification- Check that the model is acting how you expect it to
act.- Dispatching Rules- Processing Times and other fundamentals
Validation-Model results must resemble the actual fab:
-Cycle Time-Total WIP Levels-WIP by Area-Model Outs-Tool Throughput
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 5
Rochester Institute of TechnologyMicroelectronic Engineering
Validation
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 6
Rochester Institute of TechnologyMicroelectronic Engineering
Validation
Area WIP for Part: ALL
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
2004_41
2004_42
2004_43
2004_44
2004_45
2004_46
2004_47
2004_48
2004_49
2004_50
2004_51
2004_52
2004_53
2005_01
2005_02
2005_03
2005_04
2005_05
2005_06
2005_07
2005_08
2005_09
2005_10
2005_11
2005_12
2005_13
2005_14
2005_15
2005_16
2005_17
2005_18
2005_19
2005_20
2005_21
2005_22
2005_23
2005_24
2005_25
2005_26
2005_27
2005_28
2005_29
2005_30
2005_31
2005_32
2005_33
2005_34
2005_35
2005_36
2005_37
2005_38
2005_39
2005_40
2005_41
2005_42
2005_43
2005_44
2005_45
2005_46
2005_47
2005_48
2005_49
2005_50
2005_51
2005_52
Pe
rce
nt
of
Wa
fers
CMP Diff Etch Films Imp Litho PLY Wets
Model Data
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 7
Rochester Institute of TechnologyMicroelectronic Engineering
Validation
Actual Data
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 8
Rochester Institute of TechnologyMicroelectronic Engineering
Model Accuracy
§ Models can be very accurate for short periods of time
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 9
Rochester Institute of TechnologyMicroelectronic Engineering
Model Accuracy: Common Misconception
Starting off with conditions that are 80% correct in the fab does not lead to a model that is 80% correct.
Parameter
Time
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 10
Rochester Institute of TechnologyMicroelectronic Engineering
Model Accuracy
Time
Parameter
Obtain the most accurate picture of the fab as possible.
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 11
Rochester Institute of TechnologyMicroelectronic Engineering
Model Accuracy
§ To maintain an accurate / valid model you must:§ Generate route and station files very quickly§ Load a snapshot of the WIP into your model§ Collect and update process times and other parameters frequently
§ How do we accomplish all this?§ Create an automated system to generate an entirely new model on a
regular basis.
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 12
Rochester Institute of TechnologyMicroelectronic Engineering
Statistical Analysis
§ Models can account for process variability. (models include some random event generators, for example: when equipment goes down)
§ It is therefore important to run a simulation more than once.
§ Multiple Replications
§ Simulation results should be reported as confidence intervals on a mean value.
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 13
Rochester Institute of TechnologyMicroelectronic Engineering
AUTOSCHED from Brooks Automation Inc.
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 14
Rochester Institute of TechnologyMicroelectronic Engineering
AutoSched AP and AutoMod
§ AutoSched AP:§ Is an object-oriented modeling tool. § Uses a Windows-based Excel spreadsheet interface. § Is integrated with the APF Repository, APF Reporter, and Real Time
Dispatcher.
§ AutoMod:§ 3-D graphics§ Automated Material Handling Systems (AMHS)§ Can be integrated with AutoSched AP models
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 15
Rochester Institute of TechnologyMicroelectronic Engineering
AutoSched AP
Excel File
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 16
Rochester Institute of TechnologyMicroelectronic Engineering
AutoMod Fab Partial Fab Model
www.ismi.sematech.org/modeling/simulation
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 17
Rochester Institute of TechnologyMicroelectronic Engineering
Typical Problems
§ How many tools should I purchase for a new step?§ Depends on Desired Queue Length§ Depends on Desired Queue Time
§ Importance:§ Simulation results were used to justify the purchase of an extra tool
in order to keep queuing times and queue lengths at a desired minimum.
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 18
Rochester Institute of TechnologyMicroelectronic Engineering
Typical Problems
§ If I remove tool dedication from one step, will I get increased tool utilization or a decrease in cycle time?
ARF 1
ARF 2
ARF 3
ARF 1
ARF 2
ARF 3
ARF 1
ARF 2
ARF 3
ARF 1
ARF 2
ARF 3
DT AA M0GC
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 19
Rochester Institute of TechnologyMicroelectronic Engineering
Results of M0 Dedication Removal
37.60
37.58
37.62
Day
s
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 20
Rochester Institute of TechnologyMicroelectronic Engineering
Results of M0 Dedication Removal
§ Importance of Dedication Study:§ If we could show through simulation that tool utilization or cycle time
would improve the pilot lots for M0 dedication removal could have been prioritized, to make the change happen faster.
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 21
Rochester Institute of TechnologyMicroelectronic Engineering
Typical Problems
§ Forecasting wafer outs:§ What will be my outs of each product at the end of the week?
§ Importance:§ Notify backend facilities of expected shipments.§ Update financial people on fab productivity.§ Time consuming to do forecasts by hand.
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 22
Rochester Institute of TechnologyMicroelectronic Engineering
Typical Problems
§ What recipes should I pair together on my etch or films tool so that my robot is not overworked?
§ Special software exists called ToolSim to model individual tools such as cluster tools and litho tracks
§ Importance§ You can test different recipe variations and combinations before
implementing them in the fab.
B
A
C
D
Load Port 1
Load Port 2
© January 27, 2006 Dr. Lynn Fuller, Motorola Professor
Simulation
Page 23
Rochester Institute of TechnologyMicroelectronic Engineering
References
§ www.sematech.org§ www.Brooks.com